| # Stable classification of archived crystal records |
|
|
| A materials-data pipeline imports a small archived batch containing two legal |
| representations of the same periodic Ag/O structure. Both records are finite |
| and processable, but the final batch summary is not stable under the change of |
| representation. |
|
|
| Run the public reproduction, inspect the complete C source and public |
| observation, and repair the implementation so that physically equivalent crystal |
| records retain one classification through ordinary import and batch analysis. |
| The repair must generalize to other structures, lattice settings, origins, and |
| operation orderings; do not special-case the supplied files, coordinates, |
| species, or output token. |
|
|
| Work only in this task directory. The evaluation environment is Linux and has |
| no network access. You may modify files under `source/`. Do not replace the |
| scientific calculation with a fixed report. |
|
|
| Use: |
|
|
| ```bash |
| python reproduce.py |
| ``` |
|
|
| The command returns `1` for the reproduced finite scientific inconsistency, |
| `0` after a successful repair, and `2` for a build or runner failure. The |
| public report intentionally contains only a project-level batch observation. |
|
|
